Most organisations measure security culture with the one number that is easiest to collect, the phishing simulation click rate, and it tells them very little. Click rates move with the difficulty of the lure, the time of year and the channel tested, and they say nothing about whether people report problems, follow secure processes under pressure or trust the security team enough to admit a mistake. This guide sets out how to measure security culture properly: which dimensions to assess, how to design a culture survey, which behavioural and operational metrics to combine with it, and how to report results without turning them into a blame exercise. It is for security awareness leads, CISOs and human risk teams, and it goes deeper on measurement than the Security Awareness and Human Risk Management Guide.
Key takeaways
- Culture is shared values, not individual knowledge. Measure attitudes, norms and behaviours at team and organisation level; a quiz score measures knowledge, not culture.
- Combine three kinds of evidence. Perception data from surveys and interviews, behavioural data from simulations and reporting, and operational data from security systems each catch what the others miss.
- Reporting beats clicking. The share of people who report a suspicious message, and how fast, is a stronger signal of a healthy culture than the share who click.
- Control for difficulty. Rate simulation difficulty, for example with the NIST Phish Scale, or click rate trends will reflect lure design more than behaviour.
- Measure leaders and processes too. NIST CSF 2.0 GV.RR-01 makes leadership accountable for fostering a risk-aware culture, so leadership behaviour and process friction belong in the metrics.
What security culture is and why it needs measuring
The UK National Protective Security Authority (NPSA) describes security culture as the shared values that determine how people in an organisation think about and approach security. That definition matters for measurement because it puts the unit of analysis at the group, not the individual. A team where everyone passed the annual module but nobody challenges an unusual payment request has a weak culture regardless of its training scores.
The NCSC published its Cyber Security Culture Principles in June 2025. They give a usable structure for what to measure, because each principle can be turned into observable indicators:
| NCSC principle | What to measure | Example indicators |
|---|---|---|
| Frame cyber security as an enabler, supporting the organisation to achieve its goals | Whether staff see security as helping or blocking their work | Survey agreement with "security helps me do my job"; volume of security exceptions and workarounds |
| Build the safety, trust and processes to encourage openness around security | Psychological safety around reporting mistakes | Self-reported incidents (for example, "I clicked a link"); time from mistake to report |
| Embrace change to manage new threats and use new opportunities to improve resilience | Adaptability to new threats and tools | Uptake of new controls such as passkeys; speed of adoption after a threat bulletin |
| The organisation's social norms promote secure behaviours | What people believe colleagues actually do | Survey items on peer behaviour; tailgating and clear desk observation results |
| Leaders take responsibility for the impact they have on security culture | Visible leadership behaviour | Leaders' own training and MFA compliance; frequency of security topics in leadership communications |
| Provide well-maintained cyber security rules and guidelines, which are accessible and easy to understand | Policy usability | Policy read and findability tests; help desk tickets asking what a policy means |
The NCSC principles page points to NPSA's free Security Culture Tool for assessing the current shape of an organisation's security culture. Using an established instrument as a baseline, then adding organisation-specific questions, is usually better than designing a survey from scratch.
Why click rates mislead
NIST SP 800-50 Rev. 1, Building a Cybersecurity and Privacy Learning Program (September 2024), treats phishing exercises as producing two measurements: "click-through" and "reported". It advises establishing current click rates by group or work role, keeping exercises non-punitive and treating them as learning opportunities. The problems arise when click-through alone becomes the headline culture metric:
- Difficulty drives the number. A lure that imitates an internal HR process will catch far more people than a generic parcel notice. The NIST Phish Scale (NIST TN 2276, November 2023) gives a method for rating how hard an email is for a person to recognise as phishing, so results can be compared across campaigns.
- Channel blind spots. Email simulations do not measure resistance to SMS or voice attacks. In the Twilio breach, attackers texted employees outside corporate email filtering with links to fake sign-in pages, a channel that email click rates would never have measured.
- Goodhart effects. When click rate is a target, teams learn to spot the simulation platform's tells rather than real attacks, and security teams are tempted to send easier lures to show progress.
- It hides the response. One person reporting within two minutes can protect everyone else; a culture where clickers report quickly is stronger than one where fewer click but nobody reports.
Click rate still has a place as one input, normalised for difficulty and reported alongside reporting rate. For programme design, see the Phishing Simulation Programme Guide.
A measurement model: three evidence types
A credible culture assessment triangulates. SP 800-50 Rev. 1 distinguishes quantitative measurements (completion rates, closed survey items, incident reporting data, simulation tracking, longitudinal behaviour change such as timely OS updates or increased MFA adoption) from qualitative ones (attitude surveys, open responses, focus groups, interviews, case analyses). Organising them by evidence type makes gaps obvious.
| Evidence type | Sources | Strengths | Weaknesses |
|---|---|---|---|
| Perception (what people say) | Culture surveys, focus groups, interviews, exit interviews | Reveals beliefs, norms, trust and friction | Social desirability bias; survey fatigue |
| Behaviour (what people do in tests) | Phishing and vishing simulations, physical tests, tabletop exercises | Observable, comparable over time | Artificial; sensitive to test design |
| Operational (what people do for real) | Report button telemetry, help desk tickets, DLP and policy violation alerts, MFA enrolment, patch compliance, exception requests | Real behaviour at scale; no extra burden on staff | Needs careful interpretation; privacy constraints |
Metrics worth tracking
- Reporting rate: share of recipients who report a simulated or real suspicious message, by team.
- Time to first report: minutes from campaign start, or from real attack arrival, to the first user report. This is the metric that most directly limits damage.
- Self-reported mistakes: count of staff reporting their own errors (sent data to the wrong person, entered credentials on a suspicious page). A rising count after a no-blame message is a positive sign.
- Real threat report accuracy: share of user reports that turn out to be genuine threats, which shows people are reporting real attacks, not just tests.
- Secure behaviour adoption: passkey or phishing-resistant MFA enrolment, password manager use, device update compliance.
- Policy friction: number of security exceptions and shadow IT discoveries, read alongside survey items on whether security slows work.
- Difficulty-adjusted click rate: click-through grouped by Phish Scale difficulty, never reported as one blended number.
- Survey index scores: average agreement across validated items for each dimension, tracked year on year.
Designing a culture survey that produces usable data
A survey is the main source of perception data. Good design decides whether the results can be acted on.
- Use dimensions, not single questions: group three to five statements under each dimension (for example the six NCSC principles) and report the dimension score. Single items are noisy.
- Agreement scales: five or seven point Likert scales from strongly disagree to strongly agree, with a small number of reverse-worded items to detect straight-lining.
- Ask about others, not just self: "People in my team check unusual payment requests" often gets more honest answers than "I check unusual payment requests", and directly measures social norms.
- Anonymity and minimum group size: report only for groups above a threshold (often ten or more respondents) so individuals cannot be identified. State this in the invitation.
- Open questions: one or two free text items ("What makes it harder to work securely here?") produce the most actionable findings.
- Keep it short: 15 to 25 items, under ten minutes. Repeat the core items every survey so trends are valid; rotate a small set of topical questions.
- Demographic cuts: department, location, role type and tenure are usually enough. Avoid cuts that make groups too small.
Pair the survey with a small number of focus groups in teams whose scores stand out, high or low. The numbers show where; the conversations show why.
Running a culture measurement programme
- Agree purpose and governance. Decide what decisions the measurement will inform, who owns it and who sees which results. Involve HR, legal and employee representatives early, especially where works councils or data protection rules apply.
- Baseline. Run the survey, pull six to twelve months of operational data and analyse existing simulation results by difficulty. Do not change the training programme during the baseline period.
- Pilot. Test the survey with two or three teams, check completion time and item clarity, and confirm that the operational data can be extracted at team level without identifying individuals.
- Analyse by team. Look for teams where perception and behaviour disagree, for example high survey scores but low reporting. Those mismatches are where culture work pays off.
- Choose interventions per finding. Low psychological safety needs a leadership response, not more training. Policy friction needs process change. Knowledge gaps need targeted training.
- Re-measure on a cycle. Run the full survey annually and track behavioural and operational metrics monthly or quarterly. Keep the core instrument stable so trends mean something.
- Close the loop publicly. Tell staff what the survey found and what changed as a result. Response rates in the next cycle depend on it.
Rolling back is mostly about communication: if a metric or simulation causes distrust, for example a lure that imitated a bonus announcement, pause it, acknowledge the impact and remove it from trend reporting rather than defending it.
Decision guidance by organisation size
| Context | Practical approach | What to avoid |
|---|---|---|
| Small organisation (under a few hundred staff) | Short annual survey, conversations at team meetings, reporting rate and self-reported mistakes as core metrics | Team level cuts that identify individuals; expensive platforms |
| Mid-sized organisation | Established survey instrument plus custom items, quarterly operational metrics, difficulty-rated simulations | One blended click rate as the board metric |
| Large or multinational organisation | Validated survey translated and localised, regional analysis, data protection impact assessment for behavioural analytics, culture metrics in risk reporting | Individual risk scores without transparency and legal review |
| High-assurance environments (finance, critical national infrastructure) | Add insider risk and physical security behaviours, link to NPSA guidance on protective security culture | Measuring only cyber behaviours in isolation from wider security |
Reporting culture results to leadership
How results are presented decides whether they drive change or defensiveness. A useful culture report has four parts:
- Dimension scores with trend: one line per culture dimension, this year against the baseline, with the number of respondents. Avoid a single overall culture score; it hides the dimension that needs attention.
- Behaviour panel: reporting rate, time to first report and difficulty-banded click rates for the quarter, with the difficulty of each campaign stated.
- Mismatch highlights: the two or three teams where what people say and what they do diverge most, framed as questions to explore rather than league tables.
- Actions and owners: what will change as a result, who owns it and when it will be re-measured. Leadership actions, such as a senior leader talking openly about a mistake they reported, belong here as much as training changes.
Never publish team rankings that shame groups, and never show individual names in culture reporting. Individual follow-up, where justified, belongs in a separate, transparent human risk process with its own governance.
Context helps leaders read the numbers. A drop in reporting rate after a reorganisation, or a rise in exceptions after a new control, is information about the organisation, not a failure of staff. Put each metric next to the events that could explain it.
Where identity fits
Many of the strongest culture indicators are identity behaviours: whether people enrol in phishing-resistant MFA without chasing, whether they refuse to share credentials or approve unexpected MFA prompts, and whether engineers report a token or API key they accidentally committed instead of quietly deleting it. That last one is a direct measure of psychological safety in technical teams, and it has a measurable cost when it fails. Track self-reported secret exposures and time to report alongside user phishing metrics. See the Phishing-Resistant MFA Guide for the controls that reduce dependence on user vigilance.
Framework mapping
| Framework | Reference | Relevance |
|---|---|---|
| NIST CSF 2.0 | GV.RR-01 | Leadership accountable for fostering a risk-aware, ethical, continually improving culture |
| NIST CSF 2.0 | GV.RR-04 | Cybersecurity included in human resources practices |
| NIST CSF 2.0 | PR.AT-01, PR.AT-02 | Awareness and training for general and specialised roles |
| NIST SP 800-53 Rev. 5 | AT-2, AT-2(3), AT-3, AT-4 | Literacy training and awareness, social engineering and mining, role-based training, training records |
| NIST SP 800-53 Rev. 5 | PM-14 | Testing, training and monitoring activities |
| NIST SP 800-50 Rev. 1 | Measurements and metrics | Quantitative and qualitative measurement of learning programmes and culture |
| ISO/IEC 27001:2022 | Annex A 6.3 | Information security awareness, education and training |
| CIS Controls v8.1 | Control 14 | Security awareness and skills training; see the CIS Control 14 Security Awareness and Skills Training Guide |
Common failure patterns
- Reporting one blended click rate to the board. It mixes easy and hard lures, hides reporting behaviour and invites gaming.
- Punishing clickers. Disciplinary follow-up for simulation failures suppresses reporting of real incidents, which is the behaviour that matters most. SP 800-50 Rev. 1 recommends non-punitive exercises.
- Measuring only email. Attackers use SMS, voice and collaboration tools; the Twilio intrusion began with text messages to staff.
- Changing the survey every year. Rewording core items breaks trends, so improvement cannot be shown.
- Surveying without acting. Staff stop responding when nothing visibly changes after the last survey.
- Treating culture as the awareness team's job alone. Leadership behaviour and process design shape culture more than training content; metrics that exclude them miss the cause.
Practitioner checklist
- Define security culture dimensions, for example using the six NCSC principles, and map indicators to each.
- Baseline with an established instrument such as the NPSA Security Culture Tool, plus custom items.
- Rate every phishing simulation for difficulty and report click rates by difficulty band.
- Make reporting rate and time to first report the primary simulation metrics.
- Track self-reported mistakes, including accidental secret exposure, as a psychological safety signal.
- Set a minimum reporting group size and agree data handling with HR, legal and staff representatives.
- Test channels beyond email, including SMS and voice, with appropriate consent and governance.
- Include leadership behaviour and policy friction metrics in culture reporting.
- Keep core survey items stable and publish what changed after each cycle.
Standards and references
- NCSC: Cyber Security Culture Principles
- NIST: SP 800-50 Rev. 1 Building a Cybersecurity and Privacy Learning Program
- NIST: TN 2276 NIST Phish Scale User Guide
- NPSA: Security Culture
- ENISA: Cyber Security Culture in Organisations
- NIST: Cybersecurity Framework 2.0
- NIST: SP 800-53 Rev. 5
Related NHI Mgmt Group resources: Security Awareness and Human Risk Management Guide · Phishing Simulation Programme Guide · CIS Control 14 Security Awareness and Skills Training Guide · Security Awareness and Culture Manager Career Guide